Models and Policy

Few days ago, I was discussing about the opportunity for the South African Reserve Bank to publish the expected path of the policy instrument. This is a practice that is gaining support in several central banks around the world and, according to the prevalent thinking, a perfect sensible thing to do. The basic idea is very simple: because monetary policy is, in large part, about controlling private sector expectations, the Bank should make public all the informations that are necessary to make the private sector form expectations in line with the Bank predictions. Thus the Bank should say not only what she does but also what she intend to do in the near future. Simple? Not so fast. I think this is a classical case of a result that is very much dependent on the modelling framework used to generate it. Let's take, for example, Woodford prototype Neo-Keynesian model. This is a model in deviation from the steady state, where private sector expectations, in the long run, are anchored by definition to the long run properties of the model. The uncertainty is only in the dynamic path that the economy will follow after a shock. We can add some uncertainty in the parameters, but not too much. More importantly we cannot insert uncertainty affecting the steady state values of the variables (what the inflation target is, for example). In a model like this, commitment to a path of the policy instrument in response to a shock is the best way to control private sector expectations: actually, using a timeless perspective, if the system is forward looking the policy path should have a lot of inertia, i.e. should be very sticky, because the hard work is done by expectations. The policy should not interfere too much with them.
Now consider a situation like South Africa, where the private sector (and anybody else) has a lot of uncertainty about the structure of the economy, the objectives and the credibility of the Bank, let alone what the steady state of the model is. Plus the country is a small open economy and the effect of external shocks, commodity prices, financial flows reversion etc. is dramatic to say the least. In this situation the expected path of the policy instrument changes significantly at any new round of policy decision. Does communicating this path help the Bank in controlling expectations? I doubt it. On the contrary it will probably increase the confusion around objectives and efficiency of monetary policy.
In this context, credibility is about gaining reputation for competence. It is building a track record that makes the details of the policy irrelevant. It is about common definition of objectives and trust in institutions to be able to achieve them. In modern macroeconomics the complexity of the models hides a simplistic view of the constraints faced by the policy makers. Somehow, the old credibility literature (from Barro Gordon onward) was more useful in the policy context because it was more malleable; you could easily use the models to analyse a wide range of issues and train your mind accordingly. New DSGE style models instead constraint the space of analysis to a very limited set of often uninteresting issues, giving a set of very strong results with very little robustness.